Problems in the measurement of latent variables in structural equations causal models

Some problems in the measurement of latentvariables in structural equations causal models arepresented, with examples from recent empiricalstudies. Latent variables that are theoretically thesource of correlation among the empirical indicatorsare differentiated from unmeasured variablesthat are related to the empirical indicators forother reasons. It is pointed out that these shouldalso be represented by different analytical models,and that much published research has treated thisdistinction as if it had no analytic consequences.The connection between this theoretical distinctionand disattenuation effects in latent variable modelsis shown, and problems with these estimates arediscussed. Finally, recommendations are made fordecisions about whether and how to measure latentvariables when manifest variables are potentiallyavailable. Index terms: causal models, disattenuation,emergent variables, latent variable measurement,latent variables, structural equations modeling.